Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

tiny change

wenhuach219c77b1d9ab69
diff --git a/autoround/README.md b/autoround/README.md
index c2da30bd3..3146e4e53 100644
--- a/autoround/README.md
+++ b/autoround/README.md
@@ -63,6 +63,23 @@ Auto Rounding may encounter random issues with Qwen models.
ChatGlm-V1 is not supported
+## Validated Models
+LLaMAV1 done
+LLaMAv2 done
+LaMini-GPT-124M done
+QWEN1-8B done,but has random issue
+OPT-125M done
+Mistral-7b Smoke test done
+Bloom-560 smoke test done
+falcon-7b smoke test done
+gpt-leo-125m smoke test done
+stablelm-base-alpha-3b smoke test done
+dolly-v2-3b smoke test done
+mpt-7b smoke test done
+gpt-j-6b smoke test done
+chatglm2-6b smoke test done
+mixstral-7Bx8 smoke test done
+
## Reference
If you find SignRound useful for your research, please cite our paper:
```bash
diff --git a/autoround/eval/eval.py b/autoround/eval/eval.py
index 695f492f7..2ab94912b 100644
--- a/autoround/eval/eval.py
+++ b/autoround/eval/eval.py
@@ -64,7 +64,7 @@ def simple_evaluate(
batch_size=None,
max_batch_size=None,
device=None,
- no_cache=True, ##changed by wenhua
+ no_cache=True,
limit=None,
bootstrap_iters=100000,
description_dict=None,
@@ -72,7 +72,7 @@ def simple_evaluate(
decontamination_ngrams_path=None,
write_out=False,
output_base_path=None,
- lm=None ##changed by wenhua
+ lm=None
):
"""Instantiate and evaluate a model on a list of tasks.